CN101458514A - Method for detecting acceptable test data and wafer acceptable test control method - Google Patents

Abstract

The invention provides a method for detecting acceptable testing data which includes steps: collecting acceptable testing data, calculating average and variance of collected testing data; obtaining excursion tolerance corresponding to basic average, basic variance and standard width; if absolute value of difference between average of collected acceptable testing data and basic average exceeds excursion tolerance, or ratio of variance of collected acceptable testing data and basic variance exceeds excursion tolerance, collected acceptable testing data is out of control data. The invention also discloses a wafer acceptable testing control method. The method for detecting acceptable testing data and the wafer acceptable testing control method make test more accurately and reliably, and increases detecting efficiency and reduces false alarm rate.

Description

Detection can accept the test data method and wafer can be accepted test control method

Technical field

The present invention relates to detect and to accept the test data method and wafer can be accepted test control method.

Background technology

Statistical Process Control (SPC, Statistical Process Control) is one of main tool of control stabilization output in the present production run, and that uses in the quality management control of type of production enterprise is very extensive.Effectively enforcement, application SPC can in time find the problem in the production run, take suitable improvement measure, before the generation problem, and the loss that elimination problem or reduction problem are brought.

The implication of statistical Process Control (SPC) is: " use statistical technique such as control chart to come analytic process or its output, so that take measures necessary to obtain and keep the statistics state of a control, and improve process capability ".Implement SPC and be divided into two stages, the one, analysis phase, the 2nd, monitor stages.Analysis phase is after production is ready to complete, the control limit of the many groups sample data calculation control figure that collects with production run, make and analyze with control chart, histogram or carry out the process capability analysis, the check production run whether be in add up stable state and process capability whether enough.If any one can not satisfy, then seek reason, improve, and prepare again to produce and analyze, up to two purposes that reached the analysis phase, then the analysis phase can end, enter the SPC monitor stages, analyze this moment with control chart and be converted into the control control chart.The groundwork of monitor stages is to use control to monitor with control chart, wherein, the control limit of control chart is determined according to the result of analysis phase, the data of production run in time are plotted on the control chart, and close observation control chart, the fluctuation situation of control chart mid point can demonstrate controlled process or out of control, if find out of controlly, then seeks reason and eliminates its influence as early as possible.Monitoring can demonstrate fully the effect of SPC prevention and control.In the practical application of factory, all necessary for each control project through above two stages, and can repeat where necessary like this from analyzing the process of monitoring.

Having multiplely for definite method of control limit in the control chart, is to find the relevant content of more and definite control limit in 200480037968.6 the Chinese patent application at for example application number.

When using SPC in process of production at present, usually all be to calculate the variance (Sigma) that to carry out the data of quality control earlier, then according to the analysis situation of data is selected for use suitable tolerance limit coefficient, and with the product of tolerance limit coefficient and variance as control limit.And WECO rules goes for most situation as a kind of classical definite control line and the method for monitoring according to control line.WECOrules comprises following 5 rules: 1) the 3sigma control line is set, if survey to have in the data and a bit be positioned at arbitrarily outside the data area that the 3sigma control line covered, the data of this point are exactly data 2 out of control so) the 2sigma control line is set, if survey that any data are positioned at outside the 2sigma among in the data continuous 3 at 2, these data of 3 are exactly data 3 out of control so) the 1sigma control line is set, if survey among in the data continuous 5 and have be positioned at outside the 1sigma at any 4, this five point data is exactly data 4 out of control so) the 1sigma control line is set, if survey and continuous 15 points arranged in the data outside 1sigma control, be exactly data 5 out of control so) if the data of surveying have continuous 7 points to be dull on one side of center line to rise or dull decline, be exactly data out of control so.Certainly, the timeliness that it is exactly the measurement data of gathering that application WECO rules has an important prerequisite should be consistent with the timeliness of manufacturing, and measurement data acquisition because only in this way, can analyze the defective in the processing procedure in same board from the measurement data of being gathered.

And as one of them aspect of quality control, for the performance of the wafer that guarantees to offer at last the client, before wafer dispatches from the factory, also can carry out wafer for wafer and can accept test (can accept test, WaferAcceptance Test).Wafer can be accepted test and be meant wafer after finishing all processing procedures, at the testing electrical property that test structure carried out on the wafer.According to the analysis that can accept test data to wafer, also can find the unusual of silicon wafer process usually.And also used WECO rules for the quality control that can accept test data at present.Yet, find that in this process application WECO rules detects and can accept test data through the frequent warning of regular meeting's generation, and the inspection once more to accepting data afterwards finds that the data of being reported to the police are not all is data out of control.Thereby, use WECO rules and detect and can accept test data not only because frequent the warning influenced efficiency for monitoring, also not enough to some extent on the accuracy of monitoring.

Summary of the invention

The invention provides a kind of detection and can accept method and a kind of wafer of test data and can accept test control method, solve prior art and detect that can to accept test data inaccurate, the problem that efficient is lower.

For addressing the above problem, the invention provides the method that a kind of detection can be accepted test data, comprise,

Collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Obtain the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

If the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control.

Optionally, obtaining the skew tolerance corresponding with described benchmark mean value and benchmark variance comprises the following steps:

Calculate the processing procedure ability according to benchmark variance, specification width, benchmark mean value and the desired value that can accept test data, described computing formula is as follows:

C _pk＝C _p×(1-k)

$C_p=\frac{S_w}{6{\mathrm{\σ}}_0}$

$k=\frac{|{\mathrm{\μ}}_0-t|}{S_w/2}$

Wherein, C _PkBe processing procedure ability, σ ₀Be benchmark variance, S _WBe specification width, μ ₀Be benchmark mean value, t is for accepting the desired value of test data;

Obtain and the corresponding surplus of described processing procedure ability;

The difference of calculating semicircular canal lattice width and surplus is to obtain the skew tolerance.

Optionally, described surplus is more than or equal to 3 σ ₀, described σ ₀Be the benchmark variance.

Optionally, the described mean value that can accept test data of gathering calculates according to following formula:

$\stackrel{\‾}{\mathrm{\μ}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i}{n}$

Wherein, X _iBy collection can accept the data value of test data, n by collection can accept the number of test data.

Optionally, the described variance that can accept test data of gathering is calculated according to following formula:

$\mathrm{\σ}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(X_i-\stackrel{\‾}{\mathrm{\μ}})}^2}{n}}$

Wherein, X _iBy collection can accept the data value of test data.N by collection can accept the number of test data.μ by collection can accept the mean value of test data.

Optionally, described discrete tolerance is 2-5.

Optionally, also comprise, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value do not exceed the skew tolerance, and gather the variance that can accept test data and the ratio of benchmark variance does not exceed discrete tolerance, then gathering and can accepting test data is controlled data.

The present invention also provides a kind of wafer can accept test control method, comprises,

Collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Obtain the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

If the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control;

When gather and can accept to report to the police when test data is data out of control.

Optionally, obtaining the skew tolerance corresponding with described benchmark mean value and benchmark variance comprises the following steps:

Calculate the processing procedure ability according to benchmark variance, specification width, benchmark mean value and the desired value that can accept test data, described computing formula is as follows:

C _pk＝C _p×(1-k)

$C_p=\frac{S_w}{6{\mathrm{\σ}}_0}$

$k=\frac{|{\mathrm{\μ}}_0-t|}{S_w/2}$

Wherein, C _PkBe processing procedure ability, σ ₀Be benchmark variance, S _WBe specification width, μ ₀Be benchmark mean value, t is for accepting the desired value of test data;

Obtain and the corresponding surplus of described processing procedure ability;

The difference of calculating semicircular canal lattice width and surplus is to obtain the skew tolerance.

Optionally, described surplus is more than or equal to 3 σ ₀, described σ ₀Be the benchmark variance.

Compared with prior art, but the method for above-mentioned disclosed detection acceptance test data and wafer can be accepted test control method and have the following advantages: above-mentioned disclosed detection can accept the method for test data and wafer can accept test control method with benchmark mean value, the skew tolerance of benchmark variance and specification width correspondence and the discrete tolerance corresponding with the specification width are as the controlling index that detects data out of control, and described specification width generally all is to provide based on the performance requirement of product itself or obtain according to a large amount of experiment simulation results according to the client, rather than the control line that is adopted before, thereby avoided because also the inconsistent test data collection of maybe can accepting is in different process work bench the time that can accept test data collection and actual production cycle, and the inconsistent monitoring that causes of the inconsistent and process work bench on the timeliness that causes is inaccurate.

And above-mentioned disclosed detection can be accepted the method for test data and wafer, and can to accept skew tolerance in the test control method and discrete tolerance be controlling index to the contingent two kinds of ANOMALOUS VARIATIONS of data out of control in fact.Therefore, adopt skew tolerance and discrete tolerance to make that as controlling index detection is more accurate.

And, the possibility of method that above-mentioned disclosed detection can be accepted test data when calculating the skew tolerance also with reference to the test data accepted that is detected, thereby the skew tolerance that obtains tallies with the actual situation, so both guaranteed quality, also avoided owing to the frequent detection efficiency that causes of reporting to the police descends by the product that detects.

Description of drawings

Fig. 1 is that the present invention detects a kind of embodiment process flow diagram that can accept test data;

Fig. 2 is that the present invention detects the embodiment schematic diagram data that can accept test data;

Wafer of the present invention can be accepted a kind of embodiment process flow diagram of test control method during Fig. 3.

Embodiment

The present invention detects the method can accept test data and wafer and can accept test control method and provide skew tolerance and discrete tolerance as controlling index, and when calculating the skew tolerance reference data mean value, benchmark variance and corresponding to the gather specification width that can accept test data, when calculating discrete tolerance with reference to the specification width, thereby reasonably be offset tolerance and discrete tolerance.

With reference to shown in Figure 1, a kind of embodiment that the present invention detects the method that can accept test data comprises the following steps:

Step s1, collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Step s2 obtains the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

Step s3, judge the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value whether exceed the skew tolerance, or the variance that can accept test data of gathering and the ratio of benchmark variance whether exceed discrete tolerance;

Step s4, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control.

Step s5, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value do not exceed the skew tolerance, and gather the variance that can accept test data and the ratio of benchmark variance does not exceed discrete tolerance, then gathering and can accepting test data is controlled data.

Under regard to each step and be elaborated, at first for step s1, described mean value is the mean value of all data values of the test data accepted of being gathered.Make μ by collection can accept the mean value of test data, then described mean value calculation formula is as follows:

$\stackrel{\‾}{\mathrm{\μ}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i}{n}---\left(1\right)$

X wherein _iBy collection can accept the data value of test data.N by collection can accept the number of test data.

Described variance is to describe the characteristic quantity of the test data the accepted dispersion degree of being gathered.Described dispersion degree is the test data accepted of the being gathered distributed degrees with respect to the mean value of these group data.The distribution distance mean value of the test data accepted of being gathered is near more, and variance is more little; And the distribution distance mean value of the test data accepted of being gathered is far away more, and variance is big more.Make σ by collection can accept the variance of test data, then described variance computing formula is as follows:

$\mathrm{\σ}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(X_i-\stackrel{\‾}{\mathrm{\μ}})}^2}{n}}---\left(2\right)$

X wherein _iBy collection can accept the data value of test data.N by collection can accept the number of test data.μ by collection can accept the mean value of test data.

For step s2, obtain the skew tolerance corresponding and comprise the following steps: with described benchmark mean value and benchmark variance

Calculate the processing procedure ability according to benchmark variance, specification width, benchmark mean value and the desired value that can accept test data, described computing formula is as follows:

C _pk＝C _p×(1-k) (3)

$C_p=\frac{S_w}{6{\mathrm{\σ}}_0}---\left(4\right)$

$k=\frac{|{\mathrm{\μ}}_0-t|}{S_w/2}---\left(5\right)$

Wherein, C _PkBe processing procedure ability, σ ₀Be benchmark variance, S _WBe specification width, μ ₀Be benchmark mean value, t is for accepting the desired value of test data;

Obtain and the corresponding surplus of described processing procedure ability;

The difference of calculating semicircular canal lattice width and surplus is to obtain the skew tolerance.

As can be seen, the principle of described skew tolerance setting is to guarantee can also leave certain surplus under the prerequisite of operate as normal by the wafer that can accept to test under the current specification width from the step of the above-mentioned acquisition skew tolerance corresponding with described benchmark mean value and benchmark variance.

Described specification width (SPEC Width) be exactly gather the width that can accept the data area that the pairing upper specification limit of test data (USL, Up Spec Limit) and specification lower limit (LSL, Low Spec Limit) contained.Make S _WBe described specification width, the computing formula of then described specification width is as follows:

S _w＝|USL-LSL| (6)

Wherein the value of upper specification limit and specification lower limit generally is to be formulated according to the performance requirement of wafer itself or the technician obtains by a large amount of experiment simulations for wafer by the client, therefore upper specification limit and the value of specification lower limit be we can say only relevantly with the wafer self character, and come from which kind of board or haveing nothing to do of gathering when with the product data of being gathered.Thereby, adopt the specification width as the reference of calculating discrete tolerance, just can avoid that maybe can to accept test data collection inaccurate from the monitoring that different platform causes owing to the acquisition time that can accept test data and actual production time inconsistent.

Described benchmark variance and benchmark mean value are meant one section variance and the mean value that can accept test data continuously that is in steady state (SS) in the test data accepted of being gathered.Described desired value also is the optimal value of formulating according to the performance requirement of wafer itself.Described desired value is usually in the centre of upper specification limit and specification lower limit, that is: $t=\frac{\mathrm{USL}+\mathrm{LSL}}{2}.$ Wherein, USL is a upper specification limit, and LSL is the specification lower limit.Certainly, described desired value is not limited in the value of being calculated by above-mentioned formula.

| μ ₀-t| is the absolute value of the difference of benchmark mean value and desired value, has reflected the side-play amount of benchmark mean value with respect to desired value.

C in the formula (4) _pBe commonly referred to as processing procedure precision, reflected the distribution relation of the test data accepted of being gathered herein.Because concerning the test data accepted of being gathered, the specification width is determined, and the benchmark variances sigma ₀As previously mentioned, the DATA DISTRIBUTION when having reflected data stabilization and the distance of mean value, thereby C _pMore little, the benchmark variances sigma is described ₀Big more, the DATA DISTRIBUTION when also representing data stabilization and the distance of mean value are also big more.

And the k in the formula (5) is commonly referred to as process accuracy, has reflected the central tendency of the test data accepted of being gathered herein.As previously mentioned, the specification width is determined, thereby k is big more, illustrates that reference value is big more with respect to the side-play amount of desired value, has represented that also the reference value of the current test data accepted is big more with respect to the difference of desired value.

Can see C according to formula (3) _Pk=C _p* (1-k), so C _Pk≤ C _pAs previously mentioned, can accept test data, C for one group that determines _pThe DATA DISTRIBUTION when having reflected data stabilization and the distance of mean value, and k has reflected the difference of the reference value of the current test data accepted with respect to desired value.

C _pMore little, the DATA DISTRIBUTION during the expression data stabilization and the distance of mean value are also big more.K is big more, has represented that the reference value of the current test data accepted is big more with respect to the difference of desired value, (1-k) also more little.If thereby C _PkVery little, usually just reflected following several possibility: the DATA DISTRIBUTION during data stabilization and the distance of mean value are very big, or the reference value of the current test data accepted widely different with respect to desired value, or above-mentioned both are all very big, this situation just must merit attention, and the setting of skew tolerance has just become more important in a detection index.

If the skew tolerance is provided with too smallly, will cause frequent false alarm, have a strong impact on detection efficiency.And the purpose that detects originally is exactly for discovery data out of control under the restriction of the pairing specification width of data, therefore, and at this time just need be with reference to the size of specification width.Usually in order to guarantee the reliability of product wafer, when detecting, all can reserve at least 3 σ ₀Surplus, described σ ₀Be the benchmark variance.And the size of the final surplus of determining is to determine according to the processing procedure ability of above-mentioned gained.If the processing procedure ability value is very little, the so required surplus of staying just needs bigger, and if the processing procedure ability value is big, the so required surplus of staying just can be less.And for the reliability that guarantees product wafer is higher, the skew tolerance of optimizing can be got the difference of semicircular canal lattice width and surplus.Promptly be offset tolerance=S _W/ 2-margin, described margin are promptly with reference to the surplus after the processing procedure ability.Below provide the empirical value of some skew tolerances:

If C _Pk〉=2, and S _W/ 2 〉=6 σ ₀, then be offset tolerance and just be decided to be 3 σ ₀

If 1.66≤C _Pk＜2, and 5 σ ₀≤ S _W/ 2＜6 σ ₀, then be offset tolerance and just be decided to be 2 σ ₀

For step s3, the mean deviation amount of the test data accepted that the described absolute value of gathering the difference of the mean value that can accept test data and benchmark mean value has reflected the current test data accepted of gathering to be gathered when can to accept tester table stable.The absolute value of described difference of gathering the mean value that can accept test data and benchmark mean value is by | μ-μ ₀| obtain, wherein μ is exactly the mean value that can accept test data of gathering, μ ₀Be benchmark mean value.Therefore described skew tolerance is the controlling index to described mean deviation amount in fact.

Discrete tolerance then is to gather variances sigma and the benchmark variances sigma that can accept test data ₀The controlling index of ratio.The ratio of the variance that can accept test data of gathering and benchmark variance reflected the change of shape that the DATA DISTRIBUTION relative datum that can accept test data gathered distributes, described discrete tolerance is decided with reference to the specification width, in general gather the scope that DATA DISTRIBUTION and coordinate axis covered that can accept test data can not exceed benchmark distribute the scope that covered with coordinate axis 2-5 doubly, for example 2,2.5,3,3.5,4,4.5,5.

Being offset and dispersing is to have covered the contingent two kinds of ANOMALOUS VARIATIONS trend of data out of control in fact.Therefore, adopt skew tolerance and discrete tolerance to make that as controlling index detection is more accurate.

It is clearer to make that below by a concrete example the present invention detects the method that can accept test data.

Detect the method that to accept test data according to aforesaid the present invention,, calculate the average value mu and the variances sigma of the drain saturation current parameter of gathering at first as described in the step s1.

Gather drain saturation current (Idsat) parameter, the drain saturation current parameter of being gathered as shown in Figure 2.The horizontal ordinate of Fig. 2 is represented the number sequence number of the point gathered, and the drain saturation current value of the point that the ordinate representative is gathered.As can be seen from Figure 2, in the scope of the collection point of 0-2000, except an abnormity point, the data value of other points is more concentrated, can think that the collection point of 0-2000 is more stable, therefore with the data of 0-2000 as reference data.Therefore, removing after the described abnormity point can be with the mean value of the collection point of 0-2000 as the benchmark average value mu ₀, the variance yields of the collection point of 0-2000 is as the benchmark variances sigma ₀By the mathematical statistics formula can be regarded as μ ₀=516.104, σ ₀=14.82357.

Select the data of 5 lot herein, the number sequence number is that the data of collection point of 1997-2621 are as data to be detected in the corresponding diagram 2.The average value mu of described 5 lot is calculated according to formula (1), for convenience of calculation, can calculate the mean value of the drain saturation current of each lot earlier in this example ${\stackrel{\‾}{\mathrm{\μ}}}_{\mathrm{lot}}=\underset{i=1}{\overset{25}{\mathrm{\Σ}}}{\mathrm{\μ}}_{\mathrm{wafer}},$ Wherein ${\mathrm{\μ}}_{\mathrm{wafer}}=\underset{i=1}{\overset{5}{\mathrm{\Σ}}}{X}_i.$ Try to achieve μ according to described formula _Lot1=512.8944uA, μ _Lot2=514.64uA, μ _Lot3=517.4488uA, μ _Lot4=536.7584uA, μ _Lot5=511.1008uA.And then obtain the average value mu=(μ of total drain saturation current _Lot1+ μ _Lot2+ μ _Lot3+ μ _Lot4+ μ _Lot5)/5=518.56848.

Described variances sigma is calculated according to formula (2), σ ²=(X ₁-μ) ²+ (X ₂-μ) ²+ (X ₃-μ) ²+.。。。(X ₆₂₅-μ) ²/ 625, σ=16.71796 then.

As described in the step s2, obtain the skew tolerance corresponding then with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data.

Described specification width calculates by formula (6), S _W=| USL-LSL|=|745-345|=400, and desired value t=545.

Described processing procedure ability (C _Pk) calculate C according to formula (3), formula (4) and formula (5) _Pk=C _p* (1-k)=(S _W/ 6 σ ₀) * (1-2| μ ₀-t|/S _W)=(S _W/ 2-| μ ₀-t|)/3 σ ₀=(400/2-|516.104-545|)/3*14.82357=3.847566.

C _Pk2, then surplus is got 3 σ ₀, and S _W/ 2〉6 σ ₀, then be offset tolerance and just be decided to be 3 σ ₀, 3 σ ₀=44.47071.

And discrete tolerance according to the specification width, is decided to be 2.

Therefore, as | μ-μ ₀| 3 σ, promptly | μ-516.104|〉44.47071 o'clock, that is to say that the numerical range that can accept test data is at μ ₀± 3 σ ₀Outside, promptly exceeding [471.63329-560.57471] scope, the pairing test data of accepting of μ is exactly data out of control.Perhaps when the ratio of the variance of institute's image data and benchmark variance greater than 2 the time, institute's image data also is data out of control.To the data of 5 lot all in all, | μ-μ ₀|=2.46448＜44.47071, and σ/σ ₀＜2, therefore all in all the data of 5 lot are controlled.

And it can 5 lot be a detection unit in fact that the drain saturation current parameter of being gathered among Fig. 2 is detected, and wherein the 5th lot is the lot of the described detection unit of up-to-date income.

For example, with 5 above-mentioned lot, the number sequence number is that the data of collection point of 1997-2621 are as the data of 1-5lot in the corresponding diagram 2, when the reference data of 1-5lot that is detected and front more all is controlled data, reject the data of first lot, and wait for the collection of the data of next new lot, then after the data aggregation of new lot, still be one and detect unit with 5 lot, the number sequence number is that the data of collection point of 2122-2746 are as the data of 2-6lot in the corresponding diagram 2, detect unit with 2-6lot as one, detect the data of 2-6lot then, the μ of the data of 2-6lot=519.3008, μ does not exceed [471.63329-560.57471] scope, the σ of the data of 2-6lot=16.35344, σ/σ ₀＜2, then the data of 2-6lot also are controlled data.

Continue, with 5 lot of next group, in the corresponding diagram 2 the number sequence number be the data of collection point of 2247-2871 as the data of 3-7lot, detect the data of 3-7lot.The μ of the data of 3-7lot=517.4874, μ does not exceed [471.63329-560.57471] scope, the σ of the data of 3-7lot=14.4929, σ/σ ₀＜2, then the data of 3-7lot also are controlled data.

Continue, when the number sequence number is the data of collection point of 2497-3121 in detecting corresponding diagram 2, find μ=502.384 of the data of these 5 lot, μ does not exceed [471.63329-560.57471] scope, but the σ of the data of these 5 lot=32.85884, σ/σ ₀2, be data out of control so the number sequence number is the data of the collection point of 2497-3121.

Therefore that is to say that new each time detection all is 125 data removing a top lot, and add 125 data of a up-to-date lot, and with 5 lot, 625 data detect as detecting unit altogether.The rest may be inferred, constantly to institute's the detecting of drain saturation current parameter of collection to some extent.And the variance that exceeds skew tolerance or 5 lot in the difference of mean value that detects 5 lot and benchmark mean value and the ratio of benchmark variance exceed when dispersing tolerance, think that the data of 5 lot being detected are data out of control.

The method that can accept test data according to above-mentioned detection obtains a kind of wafer and can accept test control method, with reference to shown in Figure 3, comprises the following steps,

Step s10, collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Step s20 obtains the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

Step s30, judge the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value whether exceed the skew tolerance, or the variance that can accept test data of gathering and the ratio of benchmark variance whether exceed discrete tolerance;

Step s40, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control.

Step s50, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value do not exceed the skew tolerance, and gather the variance that can accept test data and the ratio of benchmark variance does not exceed discrete tolerance, then gathering and can accepting test data is controlled data.

Step s60, when gather and can accept to report to the police when test data is data out of control.

Described wafer can accept the specific operation process of test control method and method that above-mentioned detection can be accepted test data identical, and when the test data accepted that is detected was data out of control, the alert notice slip-stick artist analyzed for the test data accepted that is detected.

In sum, above-mentioned disclosed detection can accept the method for test data and wafer can accept test control method with benchmark mean value, the skew tolerance of benchmark variance and specification width correspondence and the discrete tolerance corresponding with the specification width are as the controlling index that detects data out of control, and described specification width generally all is to provide based on the performance requirement of product itself or obtain according to a large amount of experiment simulation results according to the client, rather than the control line that is adopted before, thereby avoided because also the inconsistent test data collection of maybe can accepting is in different process work bench the time that can accept test data collection and actual production cycle, and the inconsistent monitoring that causes of the inconsistent and process work bench on the timeliness that causes is inaccurate.

And above-mentioned disclosed detection can be accepted the method for test data and wafer, and can to accept skew tolerance in the test control method and discrete tolerance be controlling index to the contingent two kinds of ANOMALOUS VARIATIONS of data out of control in fact.Therefore, adopt skew tolerance and discrete tolerance to make that as controlling index detection is more accurate.

And, the possibility of method that above-mentioned disclosed detection can be accepted test data when calculating the skew tolerance also with reference to the test data accepted that is detected, thereby the skew tolerance that obtains tallies with the actual situation, so both guaranteed quality, also avoided owing to the frequent detection efficiency that causes of reporting to the police descends by the product that detects.

Though the present invention discloses as above with preferred embodiment, the present invention is defined in this.Any those skilled in the art without departing from the spirit and scope of the present invention, all can do various changes and modification, so protection scope of the present invention should be as the criterion with claim institute restricted portion.

Claims (12)

1. a detection can be accepted the test data method, it is characterized in that, comprises the following steps:

Collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Obtain the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

If the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control.

2. detection as claimed in claim 1 can be accepted the test data method, it is characterized in that, obtains the skew tolerance corresponding with described benchmark mean value and benchmark variance and comprises the following steps:

Calculate the processing procedure ability according to benchmark variance, specification width, benchmark mean value and the desired value that can accept test data;

Obtain and the corresponding surplus of described processing procedure ability;

The difference of calculating semicircular canal lattice width and surplus is to obtain the skew tolerance.

3. detection as claimed in claim 2 can be accepted the test data method, it is characterized in that, described processing procedure ability is calculated by following formula:

C _pk＝C _p×(1-k)

$C_p=\frac{S_w}{6{\mathrm{\σ}}_0}$

$k=\frac{|{\mathrm{\μ}}_0-t|}{S_w/2}$

Wherein, C _PkBe processing procedure ability, σ ₀Be benchmark variance, S _WBe specification width, μ ₀Be benchmark mean value, t is for accepting the desired value of test data.

4. detection as claimed in claim 3 can be accepted the test data method, it is characterized in that, described surplus is more than or equal to 3 σ ₀, described σ ₀Be the benchmark variance.

5. detection as claimed in claim 4 can be accepted the test data method, it is characterized in that, the described mean value that can accept test data of gathering calculates according to following formula:

$\stackrel{\‾}{\mathrm{\μ}}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i}{n}$

Wherein, X _iBy collection can accept the data value of test data, n by collection can accept the number of test data.

6. detection as claimed in claim 5 can be accepted the test data method, it is characterized in that, the described variance that can accept test data of gathering is calculated according to following formula:

$\mathrm{\σ}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(X_i-\stackrel{\‾}{\mathrm{\μ}})}^2}{n}}$

Wherein, X _iBy collection can accept the data value of test data.N by collection can accept the number of test data.μ by collection can accept the mean value of test data.

7. detection as claimed in claim 1 can be accepted the test data method, it is characterized in that, described discrete tolerance is 2 to 5.

8. detection as claimed in claim 1 can be accepted the test data method, it is characterized in that, also comprise, if the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value do not exceed the skew tolerance, and gather the variance that can accept test data and the ratio of benchmark variance does not exceed discrete tolerance, then gathering and can accepting test data is controlled data.

9. a wafer can be accepted test control method, it is characterized in that, comprises the following steps:

Collection can be accepted test data, calculates mean value and the variance that can accept test data of gathering;

Obtain the skew tolerance corresponding with benchmark mean value, benchmark variance and specification width, and the discrete tolerance corresponding with the specification width that can accept test data;

If the absolute value of difference of the mean value that can accept test data gathered and benchmark mean value exceed the skew tolerance, or gather the variance that can accept test data and the ratio of benchmark variance exceeds discrete tolerance, then gathering and can accepting test data is data out of control;

When gather and can accept to report to the police when test data is data out of control.

10. wafer as claimed in claim 9 can be accepted test control method, it is characterized in that, obtains the skew tolerance corresponding with described benchmark mean value and benchmark variance and comprises the following steps:

Calculate the processing procedure ability according to benchmark variance, specification width, benchmark mean value and the desired value that can accept test data;

Obtain and the corresponding surplus of described processing procedure ability;

The difference of calculating semicircular canal lattice width and surplus is to obtain the skew tolerance.

11. wafer as claimed in claim 10 can be accepted test control method, it is characterized in that, described processing procedure ability is calculated by following formula:

C _pk＝C _p×(1-k)

$C_p=\frac{S_w}{6{\mathrm{\σ}}_0}$

$k=\frac{|{\mathrm{\μ}}_0-t|}{S_w/2}$

Wherein, C _PkBe processing procedure ability, σ ₀Be benchmark variance, S _WBe specification width, μ ₀Be benchmark mean value, t is for accepting the desired value of test data.

12. wafer as claimed in claim 11 can be accepted test control method, it is characterized in that, described surplus is more than or equal to 3 σ ₀, described σ ₀Be the benchmark variance.

Images